Genome-Wide Association Study of Endo-Parasite Phenotypes Using Imputed Whole-Genome Sequence Data in Dairy and Beef Cattle Alan J

Genome-wide association study of endo-parasite phenotypes using imputed whole-genome sequence data in dairy and beef cattle Alan J. Twomey, Donagh P. Berry, Ross D. Evans, Michael L. Doherty, David A. Graham, Deirdre C. Purfield

To cite this version:

Alan J. Twomey, Donagh P. Berry, Ross D. Evans, Michael L. Doherty, David A. Graham, et al.. Genome-wide association study of endo-parasite phenotypes using imputed whole-genome sequence data in dairy and beef cattle. Genetics Selection Evolution, BioMed Central, 2019, 51 (1), pp.15. 10.1186/s12711-019-0457-7. hal-02105891

HAL Id: hal-02105891 https://hal.archives-ouvertes.fr/hal-02105891 Submitted on 22 Apr 2019

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Twomey et al. Genet Sel Evol (2019) 51:15 https://doi.org/10.1186/s12711-019-0457-7 Genetics Selection Evolution

RESEARCH ARTICLE Open Access Genome‑wide association study of endo‑parasite phenotypes using imputed whole‑genome sequence data in dairy and beef cattle Alan J. Twomey1,2, Donagh P. Berry1, Ross D. Evans3, Michael L. Doherty2, David A. Graham4 and Deirdre C. Purfeld1*

Abstract Background: Quantitative genetic studies suggest the existence of variation at the genome level that afects the ability of cattle to resist to parasitic diseases. The objective of the current study was to identify regions of the bovine genome that are associated with resistance to endo-parasites. Methods: Individual cattle records were available for Fasciola hepatica-damaged liver from 18 abattoirs. Deregressed estimated breeding values (EBV) for F. hepatica-damaged liver were generated for genotyped animals with a record for F. hepatica-damaged liver and for genotyped sires with a least one progeny record for F. hepatica-damaged liver; 3702 animals were available. In addition, individual cow records for antibody response to F. hepatica on 6388 genotyped dairy cows, antibody response to Ostertagia ostertagi on 8334 genotyped dairy cows and antibody response to Neospora caninum on 4597 genotyped dairy cows were adjusted for non-genetic efects. Genotypes were imputed to whole-sequence; after edits, 14,190,141 single nucleotide polymorphisms (SNPs) and 16,603,644 SNPs were available for cattle with deregressed EBV for F. hepatica-damaged liver and cows with an antibody response to a parasitic disease, respectively. Association analyses were undertaken using linear regression on one SNP at a time, in which a genomic relationship matrix accounted for the relationships between animals. Results: Genomic regions for F. hepatica-damaged liver were located on Bos taurus autosomes (BTA) 1, 8, 11, 16, 17 6 and 18; each region included at least one SNP with a p value lower than 10− . Five SNPs were identifed as signifcant (q value < 0.05) for antibody response to N. caninum and were located on BTA21 or 25. For antibody response to F. hepatica and O. ostertagi, six and nine quantitative trait loci (QTL) regions that included at least one SNP with a p value 6 lower than 10− were identifed, respectively. Gene set enrichment analysis revealed a signifcant association between functional annotations related to the olfactory system and QTL that were suggestively associated with endo-parasite phenotypes. Conclusions: A number of novel genomic regions were suggestively associated with endo-parasite phenotypes across the bovine genome and two genomic regions on BTA21 and 25 were associated with antibody response to N. caninum.

*Correspondence: [email protected] 1 Animal and Grassland Research and Innovation Centre, Teagasc, Moorepark, Fermoy, Co. Cork, Ireland Full list of author information is available at the end of the article

© The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Twomey et al. Genet Sel Evol (2019) 51:15 Page 2 of 17

Background Te objective of our study was to identify single nucle- Parasitic diseases are highly prevalent worldwide in otide polymorphisms (SNPs) associated with three endo- cattle production systems [1] and this prevalence is parasitic diseases in cattle using imputed whole-genome forecasted to increase further as a repercussion of the sequence data in a large multi-breed population. Results anticipated change in climate [2]. In Ireland, between from this study will improve our understanding of the 75.4% [3] and 83% [4] of the Irish dairy herds, show underlying genetic mechanisms of resistance in cattle to evidence of exposure to Fasciola hepatica. In addition, parasitic diseases. Tis could contribute to enhance the parasitic diseases are associated with signifcant eco- development of further control strategies for parasitic nomic losses in both dairy and beef cattle [5–7]. Details diseases, such as vaccine development, and to improve on endo-parasite diseases in cattle are in Sekiya et al. diagnostics of parasitic diseases. In addition, our results [1]. Using pedigree-based relationships between indi- provide information that would be useful in future stud- viduals, Twomey et al. [8] documented the presence of ies that attempt to increase the accuracy of genetic eval- genetic variation for two phenotypes due to F. hepatica uations, through the exploration of targeted genomic in cattle and proposed that selection and breeding ani- regions. mals resistant to F. hepatica would complement current control strategies for F. hepatica (i.e., anthelmintic Methods treatments; [9]). Genetic variability has also been doc- Te phenotypic data analysed in this study originated umented for other endo-parasites in cattle such as from two sources: (1) national abattoir cattle data con- Ostertagia ostertagi [10, 11] and Neospora caninum taining records for F. hepatica-damaged liver, and (2) [10, 12]. Since genetic gain is a function of selection serological testing of individual cows in 68 dairy herds for accuracy [13], the accuracy of traditional genetic evalu- endo-parasites. ations that exploit only pedigree relationships are hin- dered by the lack of available phenotypic data; this is Phenotypic data for F. hepatica‑damaged liver especially true for low heritability traits. Although the Individual animal records for F. hepatica-damaged study of Twomey et al. [8] was limited by the low accu- liver were available from dairy and beef cattle that were racy of the genetic evaluations, they reported a 6% unit slaughtered in 18 Irish abattoirs between February 2012 diference in the prevalence of F. hepatica-damaged and February 2018. Cattle livers were diagnosed as either liver between cows with an estimated breeding value “liver exhibits F. hepatica damage and live F. hepatica (EBV) for F. hepatica in the top 10% versus the bot- observed in the liver” or “liver exhibits F. hepatica dam- tom 10%; the mean EBV reliability of those animals was age without the identifable presence of live F. hepatica”. 0.18. It is hypothesised that the inclusion of genomic Animals were considered to have no F. hepatica-dam- information within these evaluations will improve the aged liver if none of the above diagnostic elements were reliability of the EBV for F. hepatica and thus improve detected and were slaughtered in the same abattoir and at the prediction of whether or not an animal will have F. the same date than an animal diagnosed with a F. hepat- hepatica-damaged liver. ica-damaged liver. Twomey et al. [8] described in detail Few studies have attempted to locate genomic regions the processes to generate F. hepatica-damaged liver phe- associated with endo-parasitic diseases in cattle. Using notypes, and also the procedures and criteria imposed to a genotyping panel of only 153 microsatellite markers, maximise the likelihood that only animals exposed to the Coppieters et al. [11] documented that Bos taurus auto- parasite were included in the analysis based on F. hepat- somes (BTA) 9 and 19 harboured quantitative trait loci ica-damaged liver information available from their herd- (QTL) associated with nematode burden in 768 Dutch mates. In summary, only animals with no herd movement dairy cows that were selected to be in the top 10% and after 90 days of age were retained. Animals were consid- bottom 10% of their respective sire family (n = 12) for ered as exposed if they had resided in a herd 100 days faecal egg count (FEC). Using 305 Aberdeen Angus prior to the slaughter of a herd-mate with live F. hepatica calves genotyped for 190 microsatellite markers only, present in the liver. If the slaughtered herd-mate had a F. Kim et al. [14] reported that regions on BTA8 and 12 hepatica-damaged liver with no observable live F. hepat- were associated with FEC. To the best of our knowledge, ica, an additional criterion was imposed, i.e. as well as Kim et al. [15] is the only available study that used dense having resided in the herd 100 days prior to the slaughter genome-wide SNP genotypes (31,165 SNPs after edits) of the slaughtered herd-mate, animals also had to be born to map genomic regions associated with endo-parasitic within 100 days of the birth of the slaughtered herd-mate, diseases in cattle; they reported that 12 regions in the to be considered as exposed. bovine genome contributed to signifcant genetic varia- For the purposes of our study, liver damage caused by tion in FEC. F. hepatica was dichotomized; animals were considered Twomey et al. Genet Sel Evol (2019) 51:15 Page 3 of 17

either infected (i.e., observation of F. hepatica damage in kit, and to N. caninum using the Svanovir Neospora- the liver) or not infected with F. hepatica (i.e., liver did Ab ELISA kit (Boehringer Ingelheim Svanova, Uppsala, not exhibit F. hepatica damage). Te fnal dataset con- Sweden). ELISA tests for all blood samples were carried sisted of 187,584 animals with a F. hepatica-damaged out by the same commercial laboratory (FarmLab Diag- liver record, either positive or negative. nostics, Co. Roscommon, Ireland). Antibody responses Breeding values and associated reliabilities for the to F. hepatica and O. ostertagi were expressed as optical binary trait of F. hepatica-damaged liver were estimated density ratio (ODR), and antibody response to N. cani- with an animal linear mixed model in the MIX99 genetic num was expressed as percent positivity (PP). Data edits evaluation software suite [16]. Te model and variance were as described by Twomey et al. [10]. In summary, components specifed in the genetic evaluations were the positively skewed ODR for F. hepatica was trans- those previously documented for the genetic evalua- formed using the natural logarithm to approximate a tion of F. hepatica-damaged liver [8]. Fixed efects in normally distributed variable, whereas a reciprocal trans- the model were parity/age group, stage of lactation, het- formation was used to normalize the positively skewed erosis and recombination coefcients, animal gender, PP for N. caninum. Records were removed if, at blood herd-level contemporary group, and date-by-abattoir sampling, the cow resided in a herd diferent to that in of slaughter. Te random efects were the direct addi- which it had been present at 90 days of age. Only animals ∼ 0, Aσ 2 σ 2 tive genetic efects, where a N a with a rep- with a record for antibody response to F. hepatica were A resenting the additive variance and representing the retained if they resided in a herd that had at least fve numerator relationships matrix, and a residual efect, cows positive for antibody response to F. hepatica (i.e., an ∼ 0, Iσ 2 σ 2 where e N e with e representing the residual ODR 0.4 was previously shown to lead to production I ≥ variance and representing an identity matrix. Te pedi- losses [18] and, thus, this cut-of was regarded as the pos- gree of each animal was traced back to the founder pop- itive threshold for antibody response to F. hepatica in the ulation, which was allocated to 11 genetic groups based current study) and a within-herd prevalence higher than on breed (i.e., Holstein, Friesian, Belgian Blue, Hereford, 5% with a positive antibody response to F. hepatica on the Aberdeen Angus, Simmental Limousin, Charolais, sev- day of the blood test. Since the majority of grazing cat- eral other French breeds not listed, and other breeds not tle are exposed to O. ostertagi [19, 20] and since cows in listed and unknown); the pedigree consisted of 865,173 our study were known to graze grass for most of the year, animals. Te genetic and the residual variances were all cows were considered as exposed to O. ostertagi. For equal to 0.001 and 0.115, respectively, as estimated by antibody response to N. caninum, we retained only the Twomey et al. [8] in a large subset of animals used in records as positive for cows that resided in a herd with a the present study. Only EBV for animals that had both within-herd cow prevalence of more than 1% (i.e., ELISA a genotype and an EBV reliability higher than 0.01 were test manufacturer stated that a PP ≥ 20 indicates a posi- considered for further analyses. Moreover, 16,196 ani- tive result for N. caninum, and was thus treated as such mals were retained if they had a record themselves for in our study) for antibody response to N. caninum. In F. hepatica-damaged liver (n = 10,837) or were sires that addition, the likelihood that progeny from a cow infected had at least one progeny record for F. hepatica-damaged with N. caninum will also be infected with N. caninum liver (n = 5359). Te efective record contribution (ERC) [19] is high, thus, we discarded progeny from dams that for each animal was estimated and used as the weight- had a positive result for N. caninum (86 cows were dis- ing factor for the deregression of EBV. Deregression was carded). Te fnal dataset consisted of 6892 records for undertaken using the Secant method in the MiX99 soft- antibody response to F. hepatica, 9260 records for anti- ware [17]. Deregressed EBV for animals (i.e., a combina- body response to O. ostertagi and 5289 records for anti- tion of animals with a record themselves and sires with at body response to N. caninum. least one progeny record) with an ERC lower than 1 were Each observed animal phenotype for the continuous not considered further; fnally, 3702 animals remained traits for antibody response to F. hepatica, O. ostertagi for the analyses. and N. caninum was adjusted for non-genetic efects. Genetic and non-genetic efects for antibody response Phenotypic data for antibody response to endo‑parasites to each of the three parasites were estimated using the Blood samples were collected in autumn 2016 from MIX99 software suite [16]. Te statistical model was as 10,879 dairy cows as part of a cross-sectional seroprev- previously described by Twomey et al. [10]. Fixed efects alence study of 68 Irish dairy herds. All blood samples for antibody response to each of the three parasites were were tested separately for the presence of antibodies to parity, age relative to parity median, stage of lactation, F. hepatica using the Svanovir F. hepatica-Ab ELISA kit, heterosis and recombination coefcients, and contem- to O. ostertagi using the Svanovir O. ostertagi-Ab ELISA porary group; random efects were the direct additive Twomey et al. Genet Sel Evol (2019) 51:15 Page 4 of 17

Aσ 2 σ 2 genetic efects, where a ∼ N 0, a with a represent- whole-genome sequence (WGS) was undertaken using A ing the additive variance and representing the numera- a multi-breed reference population of 2333 Bos taurus tor relationships matrix, and a residual efect, where animals from Run6.0 of the 1000 Bulls Genomes Pro- ∼ 0, Iσ 2 σ 2 e N e with e representing the residual vari- ject. All variants in the reference population were called I ance and representing an identity matrix. Te pedigree using SAMtools, and genotype calls were improved using of each animal was traced back to the founder popula- the Beagle software to provide a consensus SNP den- tion, which was allocated to 11 genetic groups based on sity across all animals. Details of the alignment to the breed (i.e., Holstein, Friesian, Belgian Blue, Hereford, UMD 3.1 Bos taurus assembly, variant calling and qual- Aberdeen Angus, Simmental Limousin, Charolais, sev- ity controls that were completed within the multi-breed eral other French breeds not listed, and other breeds not sequence reference population are as described by Daet- listed and unknown). Variance components were as pre- wyler et al. [22]. In total, 41.39 million SNPs in Run6.0 viously described by Twomey et al. [10] for the current were identifed across the genome and the average cover- dataset; the ftted genetic variance was 0.095 for antibody age (standard deviation) was 12.85 (6.94)X. Imputation of response to F. hepatica, 0.003 for antibody response to O. the HD genotypes to WGS was then completed by frst ostertagi and 0.001 for antibody response to N. caninum, phasing all HD genotypes using Eagle ([23]; version 2.3.2) while the residual variance was 0.734 for F. hepatica, and subsequently imputing to WGS using minimac3 [24]. 0.034 for O. ostertagi and 0.010 for N. caninum, which Regions with poor WGS imputation accuracy, per- resulted in heritability values of 0.12, 0.08 and 0.09, haps due to local mis-assemblies or mis-orientated con- respectively. For each cow, phenotypes for the three para- tigs, were identifed by using a larger dataset of 147,309 sites were adjusted for fxed efects (estimated with the verifed parent-progeny relationships. Mendelian errors, linear mixed model) from the cow’s respective pheno- defned as the proportion of opposing homozygotes in type. Only animals that had both an adjusted phenotype a parent-progeny pair, were estimated for each relation- and an available genotype were retained for further anal- ship and the subsequent Mendelian error rate per SNP ysis; 6388 animals remained for F. hepatica, 8334 animals was calculated. To accurately identify genomic regions remained for O. ostertagi and 4597 animals remained for with poor imputation, the R package GenWin [25] which N. caninum. fts a β-spline to the data to fnd likely infection points, was used to determine the breakpoints of genomic Genotypes regions with high Mendelian errors. Windows were Genotypes were available from six diferent genotyp- analysed using an initial window size of 5 kb and Gen- ing panels, and the number of animals genotyped per win pooled windows for which the SNP Mendelian error panel is in Additional fle 1: Table S1. In the dataset of rates were similar. Te average SNP Mendelian error rate deregressed EBV for F. hepatica-damaged liver, 87% of per window was estimated and all variants within win- the animals were genotyped using either the Illumina dows in which the mean SNP Mendelian error rate was higher than 0.02 were removed (i.e. 687,137 SNPs were bovine High-Density BeadChip (HD; n = 777,962 SNPs; 20% of animals) or the International Dairy and Beef ver- removed). Further quality control edits were also imposed on the sion 1 (IDBv1; n = 17,137 SNPs; 10% of animals), version 2 (IDBv2; n 18,004 SNPs; 32% of animals) or version imputed sequence genotypes, including the removal of = SNPs with a minor allele frequency (MAF) lower than 3 (IDBv3; n = 53,450 SNPs; 25% of animals) genotyping panels. In the dataset of adjusted phenotypes for anti- 0.002 and of SNPs that deviated from Hardy–Weinberg −6 body response to the three diferent phenotypes, most equation (P < 6 × 10 ). After these edits, 14,190,141 of the animals were genotyped on the IDBv1 (2% of ani- SNPs remained for animals with a deregressed EBV for mals), IDBv2 (76% of animals) or IDBv3 (14% of animals) F. hepatica-damaged liver, whereas 16,603,644 SNPs genotyping panel. All animals had a call rate higher than remained for cows with an antibody response to a para- 90% and only autosomal SNPs, SNPs with a reported sitic disease. chromosomal position, and SNPs with a call rate higher Te fnal dataset of animals with both a deregressed EBV than 90% were retained for each panel. for F. hepatica-damaged and an imputed sequence geno- All animals were imputed to HD using a two-step type contained 3702 animals. Te mean breed composition approach with the FImpute2 software [21]; frst, the IDB of the animals in the dataset of F. hepatica-damaged liver genotyped animals were imputed to the Bovine SNP50 was 38% Holstein–Friesian, 17% Limousin, 11% Charolais, density (50 k) and subsequently all resulting genotypes 11% Aberdeen Angus and the remaining 23% was another (including the Bovine SNP50 genotypes) were imputed breed (Fig. 1). Te fnal dataset of animals with both an to HD using a multi-breed reference population of adjusted phenotype for antibody response and an imputed 5504 HD genotyped animals. Second, imputation to sequence genotype contained 6388 animals for F. hepatica, Twomey et al. Genet Sel Evol (2019) 51:15 Page 5 of 17

0.40 reported that all had a minimal impact on the results. Te 0.35 proportions of Holstein–Friesian, Limousin, Charolais, 0.30 Simmental, Aberdeen Angus, Belgian Blue, Parthenaise, 0.25 Jersey and other breeds were ftted as covariates in the 0.20 model for F. hepatica-damaged liver; a constraint was 0.15

eed porportion imposed by setting the efect of the Hereford breed to 0.10 Br zero to avoid linear dependencies. Te genomic infation 0.05 0.00 was estimated for each association analysis and all esti- mates were less than 1.1, i.e. 1.004 for F. hepatica-damaged liver, 1.0002 for antibody response to F. hepatica, Breed 1.0133 for antibody response to O. osteragi and 1.0320 Fig. 1 The mean breed composition of animals (n 3702) in the for antibody response to N. caninum. Multiple-testing = dataset for F. hepatica-damaged liver for Aberdeen Angus (AA), correction was applied to each analysis by transforming Belgian Blue (BB), Charolais (CH), Hereford (HE), Holstein–Friesian (HO/ the p values into their corresponding q values assuming FR), Jersey (JE), Limousin (LM), Parthenaise (PT), Simmental (SI), all a false discovery rate of 5% [30]. Terefore, SNPs with a other known breeds (OTH) and unknown breed (UNK) q value lower than 0.05 were regarded as signifcant for the purpose of this study. Suggestive SNPs were defned as SNPs with a p value lower than 10−5. 8334 animals for O. ostertagi and 4597 animals for N. caninum. In the dataset of antibody response to the three parasitic diseases, the mean breed composition of the cows was Quantitative trait loci (QTL) regions 84% Holstein–Friesian, 9% Jersey with the remaining 7% Using PLINK [31], QTL regions were defned around from another breed. each SNP of interest, based on the fanking LD pattern. For each signifcant or suggestive SNP, its squared corre- 2 Association analyses lation ( r ) with all the other SNPs within 5 Mb upstream and 5 Mb downstream was calculated. Using a thresh- A genomic relationship matrix was calculated separately 2 for each of the traits using Method I of VanRaden [26] as: old of 0.5 for r , the start of the QTL region was defned ′ as the position of the SNP that was furthest upstream ZZ 2 , from the signifcant or suggestive SNP with an r higher 2 pq than 0.5; the end of the QTL region was defned as the Z where is an incidence matrix for the HD genotypes, position of the SNP that was furthest downstream with r2 centred by allele frequencies, and p and q are the domi- an with the suggestive or signifcant SNP higher than nant and recessive allele, respectively. 0.5. For QTL regions that overlapped, QTL regions were For all the traits in this study, SNP efects were esti- combined into a single QTL region and maximum and mated using linear mixed models in the WOMBAT minimum positions were set as start and end boundaries. software [27]. For each of the dependent variables, animal was included as a random efect with relationships Annotation and gene set enrichment analysis between animals accounted for via the genomic rela- Gene search using Ensembl (http://ensembl.org) on the tionship matrix. When the dependent variable was a UMD 3.1 genome build was carried out by focusing on deregressed EBV for F. hepatica-damaged liver, the phe- the defned QTL regions. In addition, protein-coding notype was weighted based on the information available. genes, which were identifed either as overlapping with Te weight was calculated as [28]: or located within QTL regions, were subject to enrich- 1 − h2 ment analysis on Te Database for Annotation, Visualiza- wi = , tion and Integrated Discovery (DAVID) v. 6.8. Functional 1−r2 c + i h2 annotations (Gene Ontology (GO) Biological Process, r2 i GO Cellular Component, GO Molecular Function and Kyoto Encyclopedia of Genes and Genomes (KEGG) where wi is the weighting factor of the ith deregressed and Reactome Pathways) were assigned to genes using 2 EBV, h is the heritability estimate for F. hepatica-dam- the functional annotation tool. P values were calculated 2 aged liver, ri is the reliability of the ith deregressed EBV, by EASE (an adoption of the Fisher Exact test to meas- and c is the proportion of genetic variance not accounted ure the gene-enrichment in annotation terms) to identify for by the SNPs and was set at 0.9. Purfeld et al. [29] functional annotations that were associated with traits of tested various values of c (i.e., 0.1, 0.2, 0.8 and 0.9) and interest. Twomey et al. Genet Sel Evol (2019) 51:15 Page 6 of 17

Results 2000 Te mean efective record contribution (ERC) was equal

s to 4.53 (ranging from 1.01 to 277.69) in the dataset of F. 1500 hepatica-damaged liver (Fig. 2).

1000 F. hepatica‑damaged liver Te associations between each SNP and F. hepatica-dam- 500

Number of record aged liver are shown on the Manhattan plot in Fig. 3. We found no SNP with a q value lower than 0.05 for F. hepat- 0 ica-damaged liver. Details of the most strongly associated 2 46810 12 14 16 18 20 >20 SNPs (with a p value < 10−6) in each QTL region for F. ERC hepatica-damaged liver are in Table 1. Six SNPs, located Fig. 2 Distribution of efective record contributions (ERC) of animals on BTA1, 8, 11, 16, 17 and 18, had a p value lower than in the dataset of deregressed F. hepatica-damaged liver EBV 10−6 (Table 1); fve of these six SNPs were classifed as

Fig. 3 Manhattan plot showing log (p values) of the association between the efect of each single nucleotide polymorphism (n 16,603,644 − 10 = SNPs) and the deregressed estimated breeding value for F. hepatica-damaged liver. The blue line is the threshold for suggestive SNPs

6 Table 1 Details of the most strongly associated SNPs (with a p value < 10− ) in each QTL region with deregressed EBV for F. hepatica-damaged liver BTA SNP name Pos (bp) p value Gene Favourable Efecta Annotation Start of QTL End of QTL Other allele (freq) genes in the QTL region

8 18 rs384464701 22,279,521 1.92 10− FTO A (0.988) 0.09 Intron 22,279,521 22,279,521 × 8 1 – 146,726,679 3.76 10− AGPAT3 G (0.998) 0.18 Intron 146,465,269 14,6811,109 PDXK, CSTB, × RRP1, TRAPPC10 7 17 rs110361447 71,554,867 1.26 10− SEC14L2 A (0.470) 0.02 Intron 71,554,867 71,554,867 × 7 16 rs800135980 14,007,063 5.27 10− – A (0.002) 0.19 Intergenic 14,007,063 14,007,063 × 7 11 rs384697649 87,844,503 7.79 10− YWHAQ T (0.996) 0.13 Intron 87,844,503 87,844,503 × 7 8 rs109859381 24,008,154 8.79 10− MLLT3 T (0.888) 0.02 Intron 24,008,154 24,008,154 × BTA: Bos taurus chromosome number, SNP: name of the single nucleotide polymorphism, pos: position of base pair on the chromosome, p value: p value of the SNP, gene: gene name in which the SNP is located, favourable allele (freq): allele with favourable efect (frequency of the favourable allele), efect: efect of the favourable allele, annotation: annotation of the SNP, start of QTL: base pair on the chromosome at the start of the QTL, end of QTL: base pair on the chromosome at the end of the QTL, other genes in the QTL region: protein-coding genes located in the QTL region that the SNP is not located in a Binary trait (0/1) Twomey et al. Genet Sel Evol (2019) 51:15 Page 7 of 17

intronic variants in one of each of the FTO, AGPAT3, regions were identifed for antibody response to F. hepat- ENSBTAG00000017404, YWHAQ, or MLLT3 genes. Te ica; the largest QTL was 7,593,750 bp long (between sixth SNP for F. hepatica-damaged liver was classifed 25.26 and 32.86 Mb) on BTA19; (see Additional fle 3: as an intergenic variant on BTA1. SNP rs384464701 on Table S3). Some suggestive SNPs were not in linkage dis- BTA18 was the most strongly associated with F. hepatica- equilibrium (LD; r2 < 0.5) with fanking SNPs, thus nine −8 damaged liver (p value = 1.92 × 10 ) and was classifed of the 37 detected QTL regions harboured only one SNP. as an intronic variant located in the FTO gene. Te fre- Fifty SNPs were suggestively associated with antibody quency of the favourable allele for each of the six SNPs response to F. hepatica in the QTL region between 80.81 with a p value lower than 10−6 ranged from 0.002 to 0.998 and 81.30 Mb on BTA16, which contained 4268 tested (Table 1). In total, 121 suggestive SNPs were identifed (p SNPs (average LD between 50 suggestive SNPs was 0.99); value < 10−5) and were located across 23 chromosomes. SNP rs434571591 had the strongest association with −6 Forty-eight QTL regions were suggestively associ- antibody response to F. hepatica (p value < 3.28 × 10 ). ated with F. hepatica-damaged liver (see Additional Five protein-coding genes were located between fle 2: Table S2). On BTA7, 21 SNPs were suggestively 80.8 and 81.30 Mb on BTA16, namely NR5A2, ENS- associated with F. hepatica-damaged liver in a single BTAG00000047428, KIF14, DDX59, and CAMSAP2. In QTL region that spanned from 65.23 to 67.30 Mb and the QTL region between 23.52 and 23.58 Mb on BTA15, this region included three protein-coding genes 43 of the 709 tested SNPs were detected as suggestively (NMUR2, GRIA1 and FAM114A2); the SNP with the associated with antibody response to F. hepatica; no gene strongest association in this region was rs475893299 was identifed in, or overlapped with, this QTL region. −6 (p value = 2.23 × 10 ). In the QTL region between 93.61 and 93.79 Mb on BTA11, 18 SNPs were sug- O. ostertagi gestively associated with F. hepatica-damaged liver; We found no SNP signifcantly associated with antibody all the genes located in this region encoded olfactory response to O. ostertagi (q value < 0.05), but nine SNPs receptors (OR1J2, OR1N1, ENSBTAG00000047728, were detected for antibody response to O. ostertagi with ENSBTAG00000047112, ENSBTAG00000046536, ENS- a p value lower than 10−6 (Table 3). Of these nine SNPs, BTAG00000045527, ENSBTAG00000045545, ENS- seven had a frequency of the favourable allele lower than BTAG00000020660, ENSBTAG00000038551, OR1Q1, 0.1, and one SNP was classifed as an intronic variant in and OR1B1); in addition, the only QTL region associated the GPA33 gene and the remaining eight as intergenic with antibody response to F. hepatica that overlapped variants. Forty-two QTL were identifed as suggestively with a QTL associated with F. hepatica-damaged liver associated with antibody response to O. ostertagi (p was on BTA11 between 93.74 and 93.89 Mb. value < 10−5; see Additional fle 4: Table S4). In total, 134 SNPs were identifed as suggestively associated with anti- Antibody response to endo‑parasites body response to O. ostertagi. Te QTL region between Te associations between each SNP and antibody response 94.02 and 98.01 Mb on BTA12 harboured 21 SNPs sug- to F. hepatica, O. ostertagi, and N. caninum are shown on gestively associated with antibody response to O. ostert- the Manhattan plots in Fig. 4. Details of the most strongly agi; no gene was located in this region. associated SNPs (with a p value < 10−6) in each QTL region for antibody response to F. hepatica, O. ostertagi, and N. N. caninium caninum are in Tables 2, 3 and 4, respectively. Five SNPs with a q value lower than 0.05 for antibody response to N. caninium were located on BTA21 (2 F. hepatica SNPs) and 25 (3 SNPs), respectively. In the QTL region We did not detect any SNP that was signifcantly associ- between 13.04 and 14.10 Mb on BTA25, one of the three ated (q value < 0.05) with antibody response to F. hepat- SNPs was classifed as an intronic variant in the PARN ica. Fifty-seven SNPs had a p value lower than 10−6 and gene and the other two as downstream variants on the were located on BTA6 (1 SNP), BTA14 (1 SNP), BTA15 micro-RNA gene, bta-mir-365-1. SNP rs133449464, that (43 SNPs), BTA18 (1 SNP) and BTA20 (11 SNPs). was classifed as an intronic variant in the PARN gene, −10 Te most highly associated SNP was rs801151638 (p was the most strongly associated (p value = 2.56 × 10 ) −8 value = 2.02 × 10 ), which was classifed as an inter- with antibody response to N. caninium (Table 4). Te genic variant on BTA14 (Table 2). Te frequency of the frequency of the favourable allele at SNP rs133449464 favourable allele ranged from 0.003 to 0.988 for the 57 was only 0.005 and none of the cows included in the SNPs with a p value lower than 10−6. In total, 226 SNPs dataset was homozygous for the favourable allele. Te were detected as suggestively associated with antibody two SNPs on BTA21 with a q value lower than 0.05 were response to F. hepatica (p value < 10−5) and 37 QTL located within the QTL between 55.07 and 56.27 Mb and Twomey et al. Genet Sel Evol (2019) 51:15 Page 8 of 17

Fig. 4 Manhattan plot showing log (p values) of the association between the efect of each single nucleotide polymorphism and antibody − 10 response to a F. hepatica, b O. ostertagi, c N. caninum. The blue line is the threshold for suggestive SNPs

classifed as intronic variants in the genes PPIP5K1 and 106.93 and 113.26 Mb) on BTA1 (see Additional fle 5: PDIA3; one of these two SNPs was also classifed as a Table S5). Almost half of the suggestive SNPs associated downstream variant in the gene C ATSPER2. with antibody response to N. caninium were located in In total, 248 suggestive and signifcant SNPs were asso- only three QTL regions, between 45.13 and 49.15 Mb ciated with antibody response to N caninium; only fve on BTA1 (24 SNPs), between 106.93 and 113.26 Mb on of these 248 SNPs had a MAF higher than 0.1. Forty-one BTA1 (66 SNPs) and between 57.83 and 58.95 Mb on QTL were identifed as associated with antibody response BTA14 (30 SNPs). to N. caninium, of which eight contained only one SNP, since the suggestive SNP was not in LD (r 2 > 0.5) with Gene set enrichment any other SNP in the fanking region of 5 Mb; the largest Of the 102 protein-coding genes, which either over- of the 41 QTL regions was 6,332,498 bp long (between lapped with, or were within QTL regions of suggestive Twomey et al. Genet Sel Evol (2019) 51:15 Page 9 of 17 - , PLK2 ENS GAPT BTAG00000026505 , SNX16 CHMP4C RALYL RAB3C , ENSBTAG00000005229, – – – , VAT1L NUDT7 Other genes in the QTL region Other genes in the QTL 5,158,840 83,212,505 22,008,166 44,643,846 67,661,877 2,3581,425 End of QTL 4,974,911 80,104,349 20,574,362 44,643,846 67,661,877 23,515,275 Start of QTL Annotation Intergenic Intergenic Intergenic Upstream Intergenic Intergenic a 0.42 0.19 0.46 0.66 0.44 0.08 Efect hepatica F. to with antibody response region ) in each QTL − 6 10 A (0.988) T (0.063) C (0.008) G (0.003) A (0.008) C (0.375) Favourable Favourable allele (freq) – – – PRR5L – – Gene − 8 − 7 − 7 − 7 − 7 − 7 × 10 × 10 × 10 × 10 × 10 × 10 p value 1.77 5.12 5.44 1.96 6.04 2.02 4,974,911 21,005,570 67,661,877 23,526,098 44,643,846 82,704,197 Pos (bp) Pos SNP name rs382942594 rs134727358 rs379765497 rs110443247 rs439064316 rs801151638 Details of the most strongly associated SNPs (with a p value < (with a p value SNPs associated Details of the most strongly Binary (0/1) trait

2 Table gene: gene name in which the SNP is located, of the SNP, p value p value: name of the single nucleotide polymorphism, pos: position of base pair on the chromosome, SNP: number, Bos taurus chromosome BTA: the start at base pair on the chromosome start of of QTL: of the SNP, annotation annotation: allele, allele), efect: efect of the favourable (frequency efect of the favourable allele with favourable allele (freq): favourable in the SNP is not located that region in the QTL genes located protein-coding region: other genes in the QTL the end of QTL, at base pair on the chromosome end of QTL: the QTL, a BTA 20 15 15 18 6 14 Twomey et al. Genet Sel Evol (2019) 51:15 Page 10 of 17 ENSBTAG00000022504, ENSBTAG00000022954, FAM78B, UCK2, TMCO1, UCK2, TMCO1, FAM78B, ENSBTAG00000022954, ENSBTAG00000022504, ALDH9A1, MGST3, LRRC52, RXRG, LMX1A, PBX1, ENSBTAG00000045987, NUF2, RGS5, RGS4 ENSBTAG00000034449, Other genes in the QTL region Other genes in the QTL MPZL1, RCSD1, CREG1, CD247, POU2F1, DUSP27, MAEL, ILDR2, TADA1, POGK, MAEL, ILDR2, TADA1, RCSD1, CREG1, CD247, POU2F1, DUSP27, MPZL1, GPLD1, MRS2, DCDC2, NRSN1, PRP4, PRP8, PRP6 SLCO5A1 SULF1, ENSBTAG00000037399, – – – – – – 6,388,604 7,584,712 7,899,389 9,801,565 34,481,525 35,740,556 70,057,265 65,072,681 62,993,200 End of QTL 945,572 7,584,712 7,866,148 9,402,063 33,029,305 65,072,681 35,375,134 62,981,078 70,057,265 Start of QTL osteratgi O. to with antibody response region ) in each QTL − 6 10 Intergenic Intergenic Intergenic Intergenic Intergenic Intergenic Intron Intergenic Intergenic Annotation a Efect 0.13 0.03 0.11 0.02 0.02 0.08 0.12 0.08 0.15 Favourable Favourable allele (freq) A (0.004) A (0.054) A (0.005) A (0.650) A (0.198) T (0.008) T (0.004) A (0.010) G (0.002) Gene – – – – – – GPA33 – – − 7 − 7 − 7 − 7 − 7 − 7 − 7 − 7 − 7 × 10 × 10 × 10 × 10 × 10 × 10 × 10 × 10 × 10 3.61 4.26 4.85 9.50 8.25 7.67 7.74 p value 4.62 5.43 9,621,581 7,893,299 1,746,600 7,584,712 33,302,807 65,072,681 62,981,078 70,057,265 35,375,134 Pos (bp) Pos Details of the most strongly associated SNPs (with a p value < (with a p value SNPs associated Details of the most strongly rs455561091 rs379990763 rs475710014 rs29026148 rs439873982 – rs209722101 SNP name rs523857350 – Binary (0/1) trait

23 12 21 4 14 3 13 3 Table BTA gene: gene name in which the SNP is located, of the SNP, p value p value: name of the single nucleotide polymorphism, pos: position of base pair on the chromosome, SNP: number, Bos taurus chromosome BTA: the start at base pair on the chromosome start of of QTL: of the SNP, annotation annotation: allele, allele), efect: efect of the favourable (frequency efect of the favourable allele with favourable allele (freq): favourable in the SNP is not located that region in the QTL genes located protein-coding region: other genes in the QTL the end of QTL, at base pair on the chromosome end of QTL: the QTL, a 14 19 Twomey et al. Genet Sel Evol (2019) 51:15 Page 11 of 17 Number of genes region in the QTL 32 9 29 18 – – 126 – 14,102,058 56,273,280 49,146,658 15,859,873 70,717,913 15,031,457 10,045,196 113,261,468 End of QTL 9,428,852 9,874,248 13,037,560 55,069,089 45,131,004 15,859,873 70,717,913 106,928,970 Start of QTL Annotation Intron Intron Intron Intron Downstream Intergenic Intergenic Intergenic a 0.09 0.11 0.08 0.03 0.07 0.03 0.02 0.07 Efect N. caninum to response with antibody ) in each QTL T (0.005) A (0.003) C (0.007) C (0.039) C (0.004) C (0.037) G (0.929) C (0.006) Favourable Favourable allele (freq) − 6 10 PARN PPIP5K1 NXPE3 VEPH1 ENSBTAG00000021175 – – – Gene − 10 − 9 − 8 − 7 − 7 − 7 − 7 − 7 × 10 × 10 × 10 × 10 × 10 × 10 × 10 × 10 p value 2.56 1.24 3.19 7.83 2.26 4.08 3.89 7.34 9,874,248 13,507,716 15,859,873 55,835,765 46,464,860 10,934,190 70,717,913 111,186,754 Pos (bp) Pos SNP name rs133449464 rs714447117 – rs466868917 rs381312356 – rs478339448 rs208471794 Details of the most strongly associated SNPs (with a p value < (with a p value SNPs associated Details of the most strongly Binary (0/1) trait

BTA 4 Table gene: gene name in which the SNP is located, of the SNP, p value p value: name of the single nucleotide polymorphism, pos: position of base pair on the chromosome, SNP: number, Bos taurus chromosome BTA: the start at base pair on the chromosome start of of QTL: of the SNP, annotation annotation: allele, allele), efect: efect of the favourable (frequency efect of the favourable allele with favourable allele (freq): favourable in the SNP is not located that region in the QTL genes located protein-coding region: other genes in the QTL the end of QTL, at base pair on the chromosome end of QTL: the QTL, a 25 9 21 1 1 3 21 5 Twomey et al. Genet Sel Evol (2019) 51:15 Page 12 of 17

SNPs for F. hepatica-damaged liver, DAVID identi- Discussion fed 97 for the gene set enrichment analysis. Te most Te globally widespread prevalence of endo-parasites in strongly associated pathways and gene ontology (GO) dairy and beef cattle is of major concern [1], since para- sets with F. hepatica-damaged liver are in Additional sitic infection is associated with reduced overall animal fle 6: Table S6. Tree pathways and 11 GO sets were performance and compromised animal health and wel- identifed with a p value lower than 0.05. Te pathway fare [32, 33]. Although there are many shortcomings of R-BTA-381753 had the strongest association with F. using anthelmintic treatments, a single annual anthel- hepatica-damaged liver and included 11 genes from the mintic treatment is still the mainstay control strategy for gene set. Te GO set with the strongest association (p F. hepatica and O. ostertagi in cattle [9, 34]. For the pre- −3 value = 9.50 × 10 ) with F. hepatica-damaged liver was vention of N. caninum infection, herd biosecurity is the the biological process of ion transmembrane transport, major control strategy adopted in herds [35]. Since sev- which included four genes from the gene set. eral studies have shown that F. hepatica, O. ostertagi, and Gene sets submitted to DAVID consisted of 324 N. caninum display genetic diversity, genetic selection genes for antibody response to F. hepatica, 333 genes for resistant cattle could be a complementary approach for antibody response to O. ostertagi, and 429 genes to current control strategies [8, 10, 11]. In addition, the for antibody response to N. caninum. In total, DAVID existence of large genetic variation for endo-parasites in recognised 307, 289, and 373 genes for antibody cattle suggests the presence of variability at the genome response to F. hepatica, O. ostertagi and N. caninum, level, which governs inter-animal variability in suscepti- respectively. The top five ranked pathways for anti- bility to endo-parasites. Information at the genome-level, body response to F. hepatica, O. ostertagi, and N. cani- pertaining to the ability of cattle to resist to parasitic num are in Additional file 7: Table S7 and the five GO infection, could help develop novel pharmaceuticals sets that were most highly associated with antibody and diagnostic tests and implement genome-enabled response to F. hepatica, O. ostertagi, and N. caninum selection. Consequently, knowledge of the biology that are in Additional file 8: Table S8. underlies such resistance to parasitic infection could help Five pathways and 28 GO sets were significantly (p reduce the prevalence of endo-parasites in cattle, and at value < 0.05) associated with antibody response to the same time reduce the reliance on anthelmintic treat- F. hepatica. The top ranked pathway and GO set for ments. Terefore, the objective of our study was to locate antibody response to F. hepatica was R-BTA-2142712 genomic regions that contribute to the ability of cat- (which included three genes from the gene set; see tle to resist infection to F. hepatica, O. ostertagi and N. Additional file 7: Table S7) and lipoxygenase pathway caninum. (included six genes from the gene set; see Additional file 8: Table S8), respectively. For antibody response to O. ostertagi, we found 14 pathways and 28 GO sets Beneft of using whole‑sequence data with a p value lower than 0.05. A strong association It is important to note that our analyses were done on was identified between the pathway olfactory trans- imputed sequence data, which may contain errors [36], −15 duction (p value = 2 × 10 ) and antibody response and thus to alleviate this issue, rare variants with a MAF to O. ostertagi; 60 genes were included in this pathway lower than 0.2%, SNPs that deviated from Hardy–Wein- and were also identified in QTL regions associated berg equilibrium (in a larger population to more conclu- with antibody response to O. ostertagi (see Additional sively identify genotyping errors and abnormalities), and file 7: Table S7). The GO biological process of G-pro- SNPs in regions with a high Mendelian error rate were tein coupled receptor signalling pathway was strongly removed, as described in the Methods section. associated with antibody response to O. ostertagi (p To evaluate the beneft of using (imputed) whole- −24 value = 5.40 × 10 ; see Additional file 8: Table S8). genome sequence in association analyses in cattle, we In the gene set enrichment analysis of antibody carried out an additional analysis in which only 50 k and response to N. caninum, 19 pathways and 23 GO sets HD SNPs were used (see Additional fle 9: Figure S1). were identified as significant (p value < 0.05). The When the associations that we detected were limited alpha-linolenic acid metabolism pathway was associ- to only SNPs on either the 50 k or HD panels, no SNP ated with antibody response to N. caninum with a p was identifed with a p value lower than 10 −6 for any of −8 value lower than 2.70 × 10 (see Additional file 7: the parasite phenotypes investigated. In addition, only Table S7). The GO metabolic function, phospholipase 4 and 2% of the QTL detected for antibody response to A2 activity, had the strongest association with anti- F. hepatica and O. ostertagi, respectively, and no QTL −7 body response to N. caninum (p value = 4.90 × 10 ; for F. hepatica-damaged liver and antibody response see Additional file 8: Table S8). to N. caninum would have been identifed if only 50 k Twomey et al. Genet Sel Evol (2019) 51:15 Page 13 of 17

SNPs were used instead of the imputed whole-genome linked to the neurological disease schizophrenia [39, sequence data. Furthermore, if we had used only the 40], asparaginase hypersensitivity [41], and migraines HD SNPs instead of the imputed whole-sequence data, [42]. In our study, we found that GRIA1 was involved only 4% of QTL reported here would have been detected in two pathways, which were over-represented for F. for F. hepatica-damaged liver and only 11, 5 and 0% of hepatica-damaged liver and are linked to the neurology QTL would have been detected for antibody response of the animal, amyotrophic lateral sclerosis, and neuro- to F. hepatica, O. ostertagi, and N. caninum, respectively. active ligand-receptor interaction. Although ectopic F. Terefore, the majority of the causal SNPs or SNPs in LD hepatica infection has been reported to cause neuro- with causal SNPs for parasitic diseases in cattle are absent logical diseases in humans [43, 44], there are no clini- on the 50 k and HD panels, which highlights the beneft cal reports of neurological diseases in cattle associated of using whole-sequence data to identify more precisely with lesions in the central nervous system. the regions of the genome associated with specifc traits. Kim et al. [15] also documented a QTL region, located It should also be noted, that our study included multi- between 48.67 and 66.21 Mb on BTA6, which was sug- ple breeds, which may not be a representative dataset of gestively associated with FEC in 584 Aberdeen Angus other studies. cattle. In our study, we detected a QTL region (contain- ing 18 suggestive SNPs) on BTA6 associated with anti- Comparison to previous genomic studies on parasitic body response to F. hepatica within the region previously diseases reported by Kim et al. [15]. We found four protein-cod- A limited number of previous studies have attempted to ing genes (UBE2 K, N4BP2, PDS5A, and RHOH) and one locate genomic regions associated with endo-parasitic RNA gene (SnRNA) that were located within this region diseases in cattle, but none used antibody response or on BTA6. Te RHOH gene likely plays a role in the deter- liver damage phenotypes to identify regions of the bovine mination of the antibody response to F. hepatica, since genome associated with parasitic diseases. In their anal- it is involved in the activation of mast cells, which are ysis of 768 Dutch dairy cows using 153 microsatellite known to increase in numbers in response to F. hepatica markers, Coppieters et al. [11] documented that FEC was infection [45]. Tere was no overlap between the asso- associated with the QTL between 22.96 and 25.23 Mb ciations identifed by Kim et al. [15] with FEC and those on BTA19 and concluded that the gene located in this identifed here for antibody response to either O. ostert- region, ITGAE, was involved in the immune response agi or N. caninum. to parasitic diseases. Although, in our study, the QTL In a study based on only four sheep that were experi- between 22.96 and 25.23 Mb on BTA19 did not include mentally infected with F. hepatica, Alvarez Rojas et al. any suggestive SNPs, a nearby region, located between [46] identifed the lipoxygenase pathway with an over- 25.26 and 32.86 Mb, contained six suggestive SNPs for representation of genes that were both up-regulated and antibody response to F. hepatica (p values ranging from down-regulated in peripheral blood mononuclear cells, −6 −6 2 weeks after infection. Tese results are supported by 6.59 × 10 to 8.09 × 10 ). Tus, one or more QTL regions on BTA19 may be associated with resistance to our fndings on the association of the lipoxygenase path- parasitic diseases. way and other pathways linked to arachidonic acid with In a study on 584 Angus cattle, Kim et al. [15] antibody response to F. hepatica. Tis was expected since reported that 12 regions on the bovine genome con- arachidonic acid, which is a polyunsaturated omega-6 tributed to signifcant genetic variation in FEC; to our fatty acid, is documented as suppressing the T1 immune knowledge, this is the only previous study that used response in mice [47, 48] that decreases during F. hepat- dense genome-wide SNP genotypes (31,165 SNPs ica infection [49]. Furthermore, studies comparing cows after edits) to map genomic regions associated with supplemented or not with a range of polyunsaturated endo-parasitic diseases in cattle. In our study, the fatty acids showed that those that were supplemented QTL between 65.23 and 67.30 Mb on BTA7 contained had higher levels of T-helper cells, T-cytotoxic cells, cells the most suggestive SNPs (21 SNPs) associated with that expressed IL-2 receptors, and CD62 adhesion mol- F. hepatica-damaged liver, and this region overlaps ecules [50], as well as a higher mean lymphocyte prolif- with one of the QTL (64.19 to 66.89 Mb) identifed by eration and higher titers of anti-OVA IgG [51]. Kim et al. [15]. One of the genes that we detected in In our study, we also expected that there would be very this region, i.e. GRIA1, was previously reported to be little overlap between QTL regions associated to the two associated with ovulation rate in 639 Japanese Black phenotypes for F. hepatica, since Twomey et al. [8] found cattle [37]. However, Cushman et al. [38] reported that only a weak genetic relationship between F. hepatica- GRIA1 had no infuence on ovulation rate in a study of damaged liver and antibody response to F. hepatica (0.37; 139 crossbred beef cows. In humans, GRIA1 has been SE = 0.28). Twomey et al. Genet Sel Evol (2019) 51:15 Page 14 of 17

Regions strongly associated with N. caninum in the GPA33 gene that was associated with a p value −7 To date, the signifcant regions associated with antibody of 7.74 × 10 ; there were 17 other suggestive SNPs response to N. caninum on BTA21 and 25 had not been identifed in this region between 0.95 and 6.39 Mb on reported to be associated with endo-parasitic diseases BTA3. Although, the frequency of the favourable allele in cattle. Te favourable alleles for the most signifcant (< 0.004) was low for all these suggestive SNPs, GPA33 SNP on BTA21 and 25 had a frequency of only 0.003 and is a likely candidate gene associated with antibody 0.005, respectively, which suggests that, if this frequency response to O. ostertagi. Indeed, in a study on mice, increased, more genetic gain is potentially possible. How- Williams et al. [59] reported that GPA33 was important ever, when we analysed exclusively the 3038 purebred for the intestinal barrier function, and thus it may have Holstein–Friesians in our study, the allele frequency of a role in the defence of gastrointestinal nematodes such the most signifcant SNP on BTA21 and 25 dropped to as O. ostertagi. 0.002 and 0.001, respectively and these SNPs were not associated with antibody response to N. caninum (p value > 0.06). Nevertheless, the most signifcant SNP on Avoidance of parasitic diseases BTA25 in the multi-breed analysis was an intronic vari- Research on animal resistance to disease focuses ant located in the PARN gene, which encodes an enzyme mainly on the immune response [60, 61]. Neverthe- (PARN) that has a role in early foetal development [52]. less, the coevolution of animals and parasites has led In cows, the gene PARN has been linked to embryo loss, to the development of animal behavioural strategies, since the level of PARN mRNA was lower in persistent including an avoidance strategy, to minimize parasite follicles (‘ageing’ follicles associated with embryo loss) infection [62]. One such avoidance strategy is avoiding than in growing follicles [53]. N. caninum infection in faecal areas caused by an odour cue, since the major- females can result in abortion or embryo loss [35] and in ity of parasites will be located in areas near (or within) reduced reproductive performance [10], thus the asso- faecal material [62]. In a study on 16 Aberdeen Angus ciation that we observed between PARN and antibody steers, Smith et al. [63] observed that cattle avoided response to N. caninum is supported by the association grazing areas contaminated by faeces, with the excep- between PARN and embryo loss. tion of rabbit faecal material. In a study on 20 sheep, Hutchings et al. [64] reported that ewes classifed as genetically resistant to parasitic diseases displayed a Identifcation of possible novel candidate genes diferent grazing behaviour compared to ewes classifed We identifed that SNP rs384464701, an intronic variant as genetically susceptible to parasitic diseases; ewes in the FTO gene on BTA18, was associated with F. hepat- that were considered as genetically resistant did not ica-damaged liver but the frequency of its favourable graze areas contaminated by faeces. As suggested by allele (0.988) was very high, thus, the potential gains for Twomey et al. [8], avoidance strategies may contribute phenotypic improvement in resistance are very limited; to the genetic diferences observed among cattle for dif- in fact, the allele is fxed in the two British breeds of the ferent parasite phenotypes. Although, avoidance tech- 301 purebred Aberdeen Angus and 164 purebred Her- niques by livestock may depend on visual clues, odour efords included in our study. Analysis of the 1248 pure- clues are most likely also involved. bred Holstein–Friesians (i.e., ≥ 87.5% Holstein–Friesian) Our fndings suggest that cattle use odour cues as in our study revealed that this SNP rs384464701 had a p part of an avoidance strategy for the three endo-para- −4 value of 3.91 × 10 and the frequency of its favourable sites studied here, since at least one pathway and GO allele was 0.003. FTO is known to have a strong asso- set involved in the olfactory system were signifcant ciation with human obesity [54]. Bravard et al. [55] sug- (p value < 0.05) for each of the parasite phenotypes. In gested that FTO has a role in the metabolic regulation of addition, the only region on the bovine genome, that the liver. In cattle, variants of the FTO gene have previ- contained QTL that were suggestively associated to F. ously been associated with both growth and carcass traits hepatica-damaged liver and antibody response to F. [56, 57] as well as with fat and protein yield in milk [58]; hepatica, included a large number of olfactory receptor taken together, these results suggest that FTO may be genes. Tis suggests that a proportion of the genetic vari- involved in energy homeostasis. However, Twomey et al. ation observed for parasitic diseases may be linked to the [10] reported that F. hepatica-damage liver had little or ability of cattle to avoid infective larvae. Nevertheless, no genetic association with milk production and carcass since genes in these suggestive regions were used in the merit in cattle. gene set enrichment analysis, further research is needed Regarding the antibody response to O. ostertagi, to support the concept that cattle avoid infection through we detected a SNP at 1,746,600 bp located on BTA3 odour cues. Twomey et al. Genet Sel Evol (2019) 51:15 Page 15 of 17

Conclusions Author details 1 Animal and Grassland Research and Innovation Centre, Teagasc, Moorepark, We identifed a large number of QTL regions that Fermoy, Co. Cork, Ireland. 2 School of Veterinary Medicine, University College were associated with the endo-parasite phenotypes Dublin, Belfeld, Dublin 4, Ireland. 3 Irish Cattle Breeding Federation, Highfeld 4 investigated in our study. We observed strong asso- House, Bandon, Co. Cork, Ireland. Animal Health Ireland, Carrick on Shannon, Co. Leitrim, Ireland. ciations between genomic regions on BTA21 and 25 with antibody response to N. caninum; the associated Acknowledgements favourable alleles had a low frequency in the stud- We gratefully acknowledge the 1000 Bull Genomes Consortium for providing accessibility to whole-genome sequence data, which were used in this study. ied population, which suggests that large genetic gain could be achieved. Competing interests The authors declare that they have no competing interests.

Availability of data and materials All data generated during this study are included in this published article. Additional fles Consent for publication Not applicable. Additional fle 1: Table S1. Number of animals with an Illumina Bovine High-Density BeadChip (HD), Illumina Bovine50 beadchip (50 k) and Ethics approval and consent to participate Low-Density BeadChip (LD) and with an International Dairy and Beef Not applicable. version 1 (V1), version 2 (V2) and version 3 (V3) genotype in the current study for deregressed EBV for F. hepatica-damaged liver, as well as for the Funding adjusted phenotype of antibody response to F. hepatica, O. ostertagi and Funding from the Irish Department of Agriculture, Food and the Marine N. caninum. STIMULUS research grants HEALTHYGENES and FLUKELESS, funding from the Science Foundation Ireland principal investigator award grant (14/IA/2576) Additional fle 2: Table S2. Chromosome number, start of quantitative and funding from a joint research grant from Science Foundation Ireland and trait locus (QTL) region, end of QTL region and number of single nucleo- the Department of Agriculture, Food and Marine on behalf of the Government tide polymorphisms (SNPs) with a p value < 1 10 5 for each QTL region − of Ireland under the Grant 16/RC/3835 (VistaMilk) are greatly appreciated. identifed as suggestively associated with F. hepatica× -damaged liver. Additional fle 3: Table S3. Chromosome number, start of quantitative trait locus (QTL) region, end of QTL region and number of single nucleo- Publisher’s Note 5 tide polymorphisms (SNPs) with a p value < 1 10− for each QTL region Springer Nature remains neutral with regard to jurisdictional claims in pub- identifed as suggestively associated with antibody× response to F. hepatica. lished maps and institutional afliations.

Additional fle 4: Table S4. Chromosome number, start of quantita- Received: 9 September 2018 Accepted: 2 April 2019 tive trait locus (QTL) region, end of QTL region and number of single 5 nucleotide polymorphisms (SNPs) with a p value < 1 10− for each QTL region identifed as suggestively associated with antibody× response to O. ostertagi. Additional fle 5: Table S5. Chromosome number, start of quantita- References tive trait locus (QTL) region, end of QTL region and number of single 1. Sekiya M, Zintl A, Doherty ML. Bulk milk ELISA and the diagnosis of para- 5 nucleotide polymorphisms (SNPs) with a p value < 1 10− for each QTL site infections in dairy herds: a review. Ir Vet J. 2013;66:14. region identifed as suggestively associated with antibody× response to N. 2. van Dijk J, Sargison ND, Kenyon F, Skuce PJ. Climate change and infectious caninium. disease: helminthological challenges to farmed ruminants in temperate regions. Animal. 2010;4:377–92. Additional fle 6: Table S6. Name, type, p value and genes for the top 5 3. Bloemhof Y, Forbes A, Danaher M, Good B, Morgan E, Mulcahy G, et al. ranked pathways and gene ontology (GO) sets for F. hepatica-damaged Determining the prevalence and seasonality of Fasciola hepatica in liver based on the EASE p value (an adoption of the Fisher Exact test to pasture-based dairy herds in Ireland using a bulk tank milk ELISA. Ir Vet J. measure the gene-enrichment in annotation terms). 2015;68:16. Additional fle 7: Table S7. Name, source, p value and number of genes 4. Selemetas N, de Waal T. Detection of major climatic and environmental for the top 5 ranked pathways for antibody response to F. hepatica, O. predictors of liver fuke exposure risk in Ireland using spatial cluster analy- ostertagi and N. caninum based on the EASE p value (an adoption of the sis. Vet Parasitol. 2015;209:242–53. Fisher Exact test to measure the gene-enrichment in annotation terms). 5. Charlier J, Claerebout E, Duchateau L, Vercruysse J. A survey to determine Additional fle 8: Table S8. Name, type, p value and genes for the top 5 relationships between bulk tank milk antibodies against Ostertagia ranked gene ontology (GO) sets for antibody response to F. hepatica, O. ostertagi and milk production parameters. Vet Parasitol. 2005;129:67–75. ostertagi and N. caninum based on the EASE p value (an adoption of the 6. Schweizer G, Braun U, Deplazes P, Torgerson PR. Estimating the fnancial Fisher Exact test to measure the gene-enrichment in annotation terms). losses due to bovine fasciolosis in Switzerland. Vet Rec. 2005;157:188–93. 7. Reichel MP, Alejandra Ayanegui-Alcérreca M, Gondim LFP, Ellis JT. What is Additional fle 9: Figure S1. Manhattan plot showing -log10(p values) of the global economic impact of Neospora caninum in cattle—the billion association between each SNP efect and F. hepatica-damaged liver for dollar question. Int J Parasitol. 2013;43:133–42. (a) 50 k data and (b) HD data. The blue line is the threshold for suggestive 8. Twomey AJ, Sayers RG, Carroll RI, Byrne N, Brien EO, Doherty ML, et al. SNPs. Genetic parameters for both a liver damage phenotype caused by Fasciola hepatica and antibody response to Fasciola hepatica phenotype in dairy and beef cattle. J Anim Sci. 2016;94:4109–19. Authors’ contributions 9. Bloemhof Y, Danaher M, Andrew F, Morgan E, Mulcahy G, Power C, et al. AJT designed the study, performed the statistical analyses, and drafted the Parasite control practices on pasture-based dairy farms in the Republic of manuscript. DCP performed the imputation, and together with DPB, partici- Ireland. Vet Parasitol. 2014;204:352–63. pated in the design of the study. DCP and DPB helped to draft the manuscript. 10. Twomey AJ, Carroll RI, Doherty ML, Byrne N, Graham DA, Sayers RG, All authors read and approved the fnal manuscript. et al. Genetic correlations between endo-parasite phenotypes and Twomey et al. Genet Sel Evol (2019) 51:15 Page 16 of 17

economically important traits in dairy and beef cattle. J Anim Sci. parasite control measures employed on Irish dairy farms. Vet Parasitol. 2018;96:407–21. 2015;207:228–40. 11. Coppieters W, Mes THM, Druet T, Farnir F, Tamma N, Schrooten C, et al. 35. Dubey JP, Schares G, Ortega-Mora LM. Epidemiology and control of Mapping QTL infuencing gastrointestinal nematode burden in Dutch neosporosis and Neospora caninum. Clin Microbiol Rev. 2007;20:323–67. Holstein–Friesian dairy cattle. BMC Genomics. 2009;10:96. 36. Brøndum RF, Ma P, Lund MS, Su G. Genotype imputation within and 12. Pan Y, Jansen GB, Dufeld TF, Hietala S, Kleton D, Lin CY, et al. Genetic across Nordic cattle breeds. J Dairy Sci. 2012;95:6795–800. susceptibility to Neospora canium infection in Holstein cattle in Ontario. J 37. Sugimoto M, Sasaki S, Watanabe T, Nishimura S, Ideta A, Yamazaki M, et al. Dairy Sci. 2004;87:3967–75. Ionotropic glutamate receptor AMPA 1 is associated with ovulation rate. 13. Rendel J, Robertson A. Estimation of genetic gain in milk yield by selec- PLoS One. 2010;5:e13817. tion in a closed herd of dairy cattle. J Genet. 1950;50:1–8. 38. Cushman RA, Miles JR, Rempel LA, McDaneld TG, Kuehn LA, Chitko-McK- 14. Kim ES, Sonstegard TS, da Silva MVG, Gasbarre LC, Van Tassell CP. Identi- own CG, et al. Identifcation of an ionotropic glutamate receptor AMPA1/ fcation of quantitative trait loci afecting gastrointestinal parasite resist- GRIA1 polymorphism in crossbred beef cows difering in fertility. J Anim ance in an experimental Angus population. Anim Genet. 2014;45:117–21. Sci. 2013;91:2640–6. 15. Kim ES, Sonstegard TS, da Silva MVG, Gasbarre LC, Van Tassell CP. 39. Choi KH, Zepp ME, Higgs BW, Weickert CS, Webster MJ. Expression Genome-wide scan of gastrointestinal nematode resistance in closed profles of schizophrenia susceptibility genes during human prefrontal Angus population selected for minimized infuence of MHC. PLoS One. cortical development. J Psychiatry Neurosci. 2009;34:450–8. 2015;10:e0119380. 40. Magri C, Gardella R, Barlati SD, Podavini D, Iatropoulos P, Bonomi S, et al. 16. Strandén I, Lidauer M. Solving large mixed linear models using precondi- Glutamate AMPA receptor subunit 1 gene (GRIA1) and DSM-IV-TR schizo- tioned conjugate gradient iteration. J Dairy Sci. 1999;82:2779–87. phrenia: a pilot case-control association study in an Italian sample. Am J 17. Strandén I, Mäntysaari EA. A recipe for multiple trait deregression. Inter- Med Genet B Neuropsychiatr Genet. 2006;141B:287–93. bull Bull. 2010;42:21–4. 41. Chen SH, Pei D, Yang W, Cheng C, Jeha S, Cox NJ, et al. Genetic variations 18. Charlier J, Hostens M, Jacobs J, Ranst B, Duchateau L, Vercruysse J. in GRIA1 on chromosome 5q33 related to asparaginase hypersensitivity. Integrating fasciolosis control in the dry cow management: the efect of Clin Pharmacol Ther. 2010;88:191–6. closantel treatment on milk production. PLoS One. 2012;7:e43216. 42. Formicola D, Aloia A, Sampaolo S, Farina O, Diodato D, Grifths LR, et al. 19. Bloemhof Y, Forbes A, Good B, Morgan E, Mulcahy G, Strube C, et al. Common variants in the regulative regions of GRIA1 and GRIA3 receptor Prevalence and seasonality of bulk milk antibodies against Dictyocaulus genes are associated with migraine susceptibility. BMC Med Genet. viviparus and Ostertagia ostertagi in Irish pasture-based dairy herds. Vet 2010;11:103. Parasitol. 2015;209:108–16. 43. Mas-Coma S, Agramunt VH, Valero MA. Neurological and ocular fasciolia- 20. Forbes AB, Vercruysse J, Charlier J. A survey of the exposure to Ostertagi sis in humans. Adv Parasitol. 2014;84:27–149. ostertagi in dairy cow herds in Europe through the measurement of anti- 44. Mas-Coma S, Valero MA, Bargues MD. Fasciola, lymnaeids and human bodies in milk samples from the bulk tank. Vet Parasitol. 2008;157:100–7. fascioliasis, with a global overview on disease transmission, epidemiol- 21. Sargolzaei M, Schenkel FS, Jansen GB, Schaefer LR. Extent of linkage dise- ogy, evolutionary genetics, molecular epidemiology and control. Adv quilibrium in Holstein cattle in North America. J Dairy Sci. 2008;9:2106–17. Parasitol. 2009;69:41–146. 22. Daetwyler HD, Capitan A, Pausch H, Stothard P, van Binsbergen R, 45. Vukman KV, Adams PN, Dowling D, Metz M, Maurer M, O’Neill SM. The Brøndum RF, et al. Whole-genome sequencing of 234 bulls facilitates efects of Fasciola hepatica tegumental antigens on mast cell function. Int mapping of monogenic and complex traits in cattle. Nat Genet. J Parasitol. 2013;43:531–9. 2014;46:858–65. 46. Alvarez Rojas CA, Scheerlinck JP, Ansell BRE, Hall RS, Gasser RB, Jex AR. 23. Loh PR, Danecek P, Palamara PF, Fuchsberger C, Reshef YA, Finucane HK, Time-course study of the transcriptome of peripheral blood mononu- et al. Reference-based phasing using the Haplotype Reference Consor- clear cells (PBMCs) from sheep infected with Fasciola hepatica. PLoS One. tium panel. Nat Genet. 2016;48:1443–8. 2016;11:e0159194. 24. Das S, Forer L, Schönherr S, Sidore C, Locke AE, Kwong A, et al. Next- 47. Wallace FA, Miles EA, Evans C, Stock TE, Yaqoob P, Calder PC. Dietary fatty generation genotype imputation service and methods. Nat Genet. acids infuence the production of Th1- but not Th2-type cytokines. J 2016;48:1284–7. Leukoc Biol. 2001;69:449–57. 25. Beissinger TM, Rosa GJ, Kaeppler SM, Gianola D, De Leon N. Defning 48. Monk JM, Turk HF, Fan YY, Callaway E, Weeks B, Yang P, et al. Antagonizing window-boundaries for genomic analyses using smoothing spline tech- arachidonic acid-derived eicosanoids reduces infammatory Th17 and niques. Genet Sel Evol. 2015;47:30. Th1 cell-mediated infammation and colitis severity. Mediat Infamm. 26. VanRaden PM. Efcient methods to compute genomic predictions. J 2014;2014:917149. Dairy Sci. 2008;91:4414–23. 49. O’Neill SM, Brady MT, Callanan JJ, Mulcahy G, Joyce P, Mills KH, et al. 27. Meyer K, Tier B. “SNP Snappy”: a strategy for fast genome-wide association Fasciola hepatica infection downregulates Th1 responses in mice. Parasite studies ftting a full mixed model. Genetics. 2012;190:275–7. Immunol. 2000;22:147–55. 28. Garrick DJ, Taylor JF, Fernando RL. Deregressing estimated breeding val- 50. Gandra JR, Barletta RV, Mingoti RD, Verdurico LC, Freitas JE Jr, Oliveira LJ, ues and weighting information for genomic regression analyses. Genet et al. Efects of whole faxseed, raw soybeans, and calcium salts of fatty Sel Evol. 2009;41:55. acids on measures of cellular immune function of transition dairy cows. J 29. Purfeld DC, Bradley DG, Evans RD, Kearney FJ, Berry DP. Genome-wide Dairy Sci. 2016;99:4590–606. association study for calving performance using high-density genotypes 51. Caroprese M, Marzano A, Entrican G, Wattegedera S, Albenzio M, Sevi A. in dairy and beef cattle. Genet Sel Evol. 2015;47:47. Immune response of cows fed polyunsaturated fatty acids under high 30. Storey JD, Base AJ, Dabney A, Robinson D. qvalue: Q-value estimation for ambient temperatures. J Dairy Sci. 2009;92:2796–803. false discovery rate control. R package version 2.10.0. 2015. http://githu 52. Godwin AR, Kojima S, Green CB, Wilusz J. Kiss your tail goodbye: the role b.com/jdstorey/qvalue. Accessed 13 Feb 2019. of PARN, Nocturnin, and Angel deadenylases in mRNA biology. Biochim 31. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. Biophys Acta. 2013;1829:571–9. PLINK: a tool set for whole-genome association and population-based 53. Lingenfelter BM, Dailey RA, Inskeep EK, Vernon MW, Poole DH, Rhinehart linkage analyses. Am J Hum Genet. 2007;81:559–75. JD, et al. Changes of maternal transcripts in oocytes from persistent fol- 32. Mezo M, González-Warleta M, Castro-Hermida JA, Muiño L, Ubeira FM. licles in cattle. Mol Reprod Dev. 2007;74:265–72. Association between anti-F. hepatica antibody levels in milk and produc- 54. Fawcett KA, Barroso I. The genetics of obesity: FTO leads the way. Trends tion losses in dairy cows. Vet Parasitol. 2011;180:237–42. Genet. 2010;26:266–74. 33. Charlier J, De Cat A, Forbes A, Vercruysse J. Measurement of antibodies 55. Bravard A, Vial G, Chauvin MA, Rouillé Y, Bailleul B, Vidal H, et al. FTO con- to gastrointestinal nematodes and liver fuke in meat juice of beef cattle tributes to hepatic metabolism regulation through regulation of leptin and associations with carcass parameters. Vet Parasitol. 2009;166:235–40. action and STAT3 signalling in liver. Cell Commun Signal. 2014;12:4. 34. Selemetas N, Phelan P, O’Kiely P, de Waal T. The efects of farm man- 56. Jevsinek Skok D, Kunej T, Kovac M, Malovrh S, Potocnik K, Petric N, et al. agement practices on liver fuke prevalence and the current internal FTO gene variants are associated with growth and carcass traits in cattle. Anim Genet. 2016;47:219–22. Twomey et al. Genet Sel Evol (2019) 51:15 Page 17 of 17

57. Rempel LA, Casas E, Shackelford SD, Wheeler TL. Relationship of polymor- 61. McManus C, do Prado Paim T, de Melo CB, Brasil BS, Paiva SR. Selection phisms within metabolic genes and carcass traits in crossbred beef cattle. methods for resistance to and tolerance of helminths in livestock. Para- J Anim Sci. 2012;90:1311–6. site. 2014;21:56. 58. Zielke LG, Bortfeldt RH, Reissmann M, Tetens J, Thaller G, Brockmann GA. 62. Hart BL. Behavioral adaptations to pathogens and parasites: fve strate- Impact of variation at the FTO locus on milk fat yield in Holstein dairy gies. Neurosci Biobehav Rev. 1990;14:273–94. cattle. PLoS One. 2013;8:e63406. 63. Smith LA, White PCL, Marion G, Hutchings MR. Livestock grazing behavior 59. Williams BB, Tebbutt NC, Buchert M, Putoczki TL, Doggett K, Bao S, et al. and inter- versus intraspecifc disease risk via the faecal–oral route. Behav Glycoprotein A33 defciency: a new mouse model of impaired intestinal Ecol. 2009;20:426–32. epithelial barrier function and infammatory disease. Dis Model Mech. 64. Hutchings MR, Knowler KJ, McAnulty R, McEwan JC. Genetically resistant 2015;8:805–15. sheep avoid parasites to a greater extent than do susceptible sheep. Proc 60. Råberg L, Graham AL, Read AF. Decomposing health: tolerance and Biol Sci. 2007;274:1839–44. resistance to parasites in animals. Philos Trans R Soc Lond B Biol Sci. 2009;364:37–49.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions